Three experimental tasks explored young children's metalinguistic awareness of word boundaries. First, preschool children recited word-by-word pieces which differed in abstractness and familiarity of vocabulary. Second, a teaching task was employed to train and then sample selected items taken from the recited pieces, representing words with cards. The data suggested that several strategies were employed: segment-by-phrase, segment-by-syllable, and segment-by-word. Children were less successful in segmenting the more abstract piece than the concrete one, and they tended to revert to a phrase strategy and/or to change abstract words into more common vocabulary. Error analysis revealed growing awareness of the functors a and the. In a third experiment, preschool and kindergarten children segmented phrases using the functors a, an, my, or and phrases using phonetically similar embedded syllables (e.g., hold my nose; gold miner). Segmentation scores increased with age; even the 4½- to 5-year-olds were highly successful, segmenting with 60% accuracy, and by age 6–6½, they were correctly segmenting 75% of the phrases. Children who had begun to read performed better overall than their same-age peers who had not. Children were more successful with my, or, and word-medial a than with an and word-initial a; embedded syllables tended to be easier than functors, especially for the younger subjects. In conclusion, the years from age 4½- to 6½ appear to be a period of vigorous development of word segmentation skills. During this period, children use increasingly complex strategies – first phrasal, then syllabic, and finally a full word strategy – and they demonstrate a growing knowledge of function words as well as content words. Functors vary in difficulty and developmental sequence due to differences in stress, amount of ambiguity with phonetically similar embedded syllables, semantic complexity, and salience of reference to the young child.